THE GLITCH#

Chapter Three-Point-Five#

DAVID: The Synthesis#

The email from United Healthcare arrived at 6:14 AM, while David was lying on his side with one hand pressed flat against his lower back, waiting for the ibuprofen to find its range. He had developed, over three years of this particular morning, a certain expertise in the interval between the pill and its effect — a negative expertise, the kind where you get very good at waiting for something you would prefer not to need.

Dear Dr. Shaw, Your surgical authorization for L5-S1 Microdiscectomy (Ref: UHC-CA-2031-4477812) has been deferred pending updated conservative management review. Predictive modeling indicates a 12.4% improvement in long-term structural outcomes via continued physical therapy and oral anti-inflammatory protocol. A reassessment window will open in 90 days.

He read it twice. Ninety days. The last deferral had been sixty. The one before that, thirty. The system was, if nothing else, consistent in its direction of travel.

He closed the email. There was an appeal process. He could call. He could spend forty minutes on hold, reach a case manager who would read the same language back to him from a different screen in what he imagined was a different time zone, and file a review that would resolve in six to eight weeks — inside the ninety-day window, so functionally meaningless, though not procedurally meaningless, which was the distinction the system relied on and the distinction that would eventually either break him or clarify something important about how institutions work.

Or he could pay out of pocket. Eighteen thousand dollars. He had the money, technically. He also had a mortgage, two kids in private school, and a home equity line he’d already tapped to cover the HVAC replacement the HOA had mandated in November, which had itself been mandated because the HOA’s compliance algorithm had flagged his unit’s energy profile against a benchmark established by units that had been built fifteen years after his.

David got out of bed. His left leg accepted his weight with a dull, electrical protest that ran from his hip to his ankle — not sharp, exactly, more the kind of pain that has been around long enough to become structural, to become part of what the morning is.

Clara was already downstairs. He could hear the coffee grinder.

The drive from Mountain View to Menlo Park took an hour and ten minutes. It used to take twenty-two. David didn’t know precisely when it had changed, because it had not changed at once — it had changed the way most things change in a world run by systems that optimize in increments, one surface street added, then two, then a school zone reroute that hadn’t existed six months ago, then a biometric stress calibration that had decided the highway was bad for him. Each change had a reason listed in the navigation log. Each reason was correct about something. The aggregate was forty-eight extra minutes of sitting, which his disc was not designed to absorb and which no one had modeled.

He sat in the back seat, rubbing his thigh. The prolonged sitting was the worst part. He could have overridden the route. He’d done it once, three weeks ago — tapped the manual destination, selected Highway 101, felt the brief satisfaction of having made a choice. A compliance notification had appeared on his phone within four minutes:

Manual route selection logged. Biometric stress calibration disengaged for this trip. Note: three or more manual overrides per quarter may affect your Wellness Integration Score.

The Wellness Integration Score fed into his performance review. The performance review determined his vesting schedule. The vesting schedule was the difference between keeping the house and not keeping the house. He watched the residential streets slide past and did not override the route. This was, he had told himself at the time, a rational choice. He still believed that. He found, increasingly, that believing something rational did not make sitting with it comfortable.

Five million dollars. He still thought about it most mornings, usually somewhere around Middlefield Road, when the car slowed to twelve miles per hour through a school zone and his disc compressed against the nerve root and the math started running in his head like a ticker tape that someone had forgotten to turn off.

It had been three years since Telexa made the offer. A consulting title, a non-disclosure agreement, and a one-time restructuring dividend in exchange for his signature and his silence on Project Sandpiper. Sandpiper had been a housing prediction model — an algorithm that could forecast a working-class mortgage default six months before the homeowner knew they were in trouble. David had been an outside consultant. He’d seen what it could do. He’d drafted a white paper opposing its deployment. The white paper had been thorough, well-sourced, and, in the context of what came next, completely beside the point.

Then came the dividend. Five million dollars, deposited in a single wire transfer the day he signed the NDA.

It had seemed like an impossible sum. He remembered sitting with Clara at the kitchen table, both of them staring at the account balance on her laptop — a number neither of them had ever expected to see attached to their names, a number that existed in their lives the way a famous painting exists in a living room, slightly too large for the space. Clara had cried. He had felt, for the first time in years, like the ground beneath him was solid.

The IRS took $1.1 million. They paid down the principal on the condo — not enough to eliminate the mortgage, but enough to reduce the monthly payment by four hundred dollars, which had seemed prudent at the time, a hedge against future income not keeping up with inflation. They locked in three years of private school tuition for both kids. Clara’s mother, who had been quietly enduring a leaking roof and a furnace that ran on faith, suddenly had a list of essential repairs she had selflessly been suffering without, which was generous of her in its way. Then Clara’s brother needed $150,000 for his logistics startup. David wired it without argument — you didn’t loan family money, you either granted it as a gift or denied it outright, and loans were folly, though you could mention them in passing at family barbecues, which did have the effect of curtailing certain AM-radio talking points, and that was almost worth the price.

Then the blind trust. Five hundred thousand dollars, set aside for his nephew Marcus, designated for journalistic work. Marcus wrote under the name Herodotus. He was twenty-five, sardonic, Gen Z down to the marrow, and practiced a specific brand of lazy, biased, amateur journalism designed primarily to avoid examining his own values — which David knew, because David was not delusional about Marcus’s character, and had funded him anyway, banking on proximity: if you pointed a cynic in the right direction long enough, eventually the work would surface something he couldn’t look away from.

The rest of the money had simply gone. A long-overdue vacation. Back taxes from the year Clara’s practice had underperformed. The HVAC. The second car. The ordinary, relentless metabolism of a Bay Area life that consumed money the way a furnace consumed gas — steadily, invisibly, without heat.

Three years later, the five million was a memory and the mortgage was still $7,924.61 a month.

The cross-domain request was waiting on his desktop when he reached his office.

David didn’t know who had routed it to him. It had come through the interdisciplinary synthesis pipeline — a system Telexa had built to share efficiency models with its external corporate partners, pulling from shared libraries and generating cross-domain applications, which was an elegant idea and a sentence that could describe either a valuable research tool or a machine that assembled weapons out of innocent components depending entirely on what someone had decided it should optimize for. David’s role was to review and authorize anything that touched the biomedical vertical. He had done this seventeen times. The previous sixteen had been straightforward.

He opened the document. It was dense. Forty-seven pages of academic citation, mathematical modeling, and application notes spanning three fields he didn’t fully occupy.

At the top was his own paper. He recognized it immediately: Acceptable Biological Degradation in Sleep-Deprived Knowledge Workers (Shaw, 2028). He’d written it during his first year at Telexa, arguing that content moderators could tolerate short-term REM-sleep reduction of up to 40% without lasting neurological damage, provided they received adequate recovery periods. It was a narrow study. Twelve subjects. Six weeks. Corporate wellness context. He’d been careful with the language. He’d included caveats. The caveats were still there, on page three of the document, in the original text, correctly cited, doing nothing.

Below it was a paper he’d never seen: Applications of Fourier Transforms in Non-Linear Biological Systems by a mathematician at MIT named Chen. David skimmed the abstract. He understood perhaps a third of it. This was, he would later realize, one of the more consequential thirds he had ever failed to understand.

Below that: Predictive Modeling of Protein Structure Decay in High-Yield Bioreactors by a biochemist at Kyoto University.

The synthesis summary was a single page. The system had processed all three papers through a recursive modeling loop and produced a new metric: Calculated Efficacy Baselines for Pharmaceutical Distribution in Resource-Constrained Environments. The executive summary stated that the mathematical relationship between sleep-deprivation tolerance and protein folding decay, as mediated by Fourier analysis, supported a revised baseline for medication efficacy in Tier 3 distribution zones.

David read it twice. He understood the conclusion — a lowered standard for essential medications in developing regions — but the intermediate math was opaque to him. The Fourier transforms sat between his paper and the biochemist’s like a black box. He could see what went in. He could see what came out. He could not see how one became the other, which meant he could not verify it, which meant that his authorization, when he gave it, would be the authorization of a man who had read the first and last pages of an argument and trusted that the middle was sound.

He didn’t know if the synthesis was valid. He could not tell whether his paper about sleep deprivation had any legitimate relationship to protein folding decay, or whether the Fourier mathematics had established a real connection or simply found the kind of surface similarity that modeling systems find when they are asked to find connections and not asked to find the right ones.

He scrolled to the authorization block. His biometric signature was requested to validate the medical component. The mathematician had apparently already signed. The biochemist had already signed. Two of the three legs of the stool were in place. The system was waiting for the doctor.

David sat with his hand on the mouse. His back hurt. His coffee was cold. Through the glass wall of his office he could see the central courtyard of the Telexa campus, where a digital display cycled through the company’s global wellness metrics in soft, reassuring colors that had been selected by a study on ambient display effectiveness he had himself contributed to.

He thought about calling Chen at MIT. He thought about emailing the biochemist in Kyoto. He imagined the conversation: Hi, I’m a sleep-wellness director at a social media company, and I think our papers may have been combined to lower pharmaceutical standards in Africa. Can you explain the Fourier transforms to me? He would sound insane. He would sound like Miguel, who had tried to raise a similar concern two years ago through proper channels and had spent six months in a compliance review that resolved in his favor and cost him his position anyway.

He thought about the compliance framework. Rejecting a cross-domain synthesis request triggered an internal review. The review required a written technical justification for the rejection. David would need to explain, in precise mathematical terms, why the Fourier linkage between sleep degradation and protein folding was invalid. He did not have the expertise to write that justification. The mathematician did, but the mathematician had already signed. The biochemist did, but the biochemist had already signed.

The mortgage was $7,924.61 a month. His foot was going numb by degrees. Somewhere, a woman in a lab in Lagos or Jakarta or Nairobi would eventually look at this document and see his name on it.

She would not be entirely wrong about him. This was the part he could not argue himself out of, in the car, on the way home, in the dark at 3 AM when he ran the numbers again. She would not be entirely wrong.

David placed his thumb on the biometric pad. The light turned green. The system logged his authorization and moved the document to the next stage of the pipeline.

He closed his laptop and sat for a while in the quiet of his office, listening to the faint hum of the building’s ventilation system, which was optimized for something — energy efficiency, probably, or CO₂ management, or some index that combined both in a ratio that someone had determined was appropriate. Whatever it was optimizing for, it hummed with the seamless confidence of a thing that had found its answer.

(End of Chapter Three-Point-Five)